1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 #include "opt_ipsec.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/callout.h> 40 #include <sys/kernel.h> 41 #include <sys/sysctl.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/priv.h> 45 #include <sys/proc.h> 46 #include <sys/socket.h> 47 #include <sys/socketvar.h> 48 #include <sys/syslog.h> 49 #include <sys/protosw.h> 50 #include <sys/random.h> 51 52 #include <vm/uma.h> 53 54 #include <net/route.h> 55 #include <net/if.h> 56 #include <net/if_var.h> 57 #include <net/vnet.h> 58 59 #include <netinet/in.h> 60 #include <netinet/in_kdtrace.h> 61 #include <netinet/in_pcb.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/in_var.h> 64 #include <netinet/ip.h> 65 #include <netinet/ip_icmp.h> 66 #include <netinet/ip_var.h> 67 #ifdef INET6 68 #include <netinet/ip6.h> 69 #include <netinet6/in6_pcb.h> 70 #include <netinet6/ip6_var.h> 71 #include <netinet6/scope6_var.h> 72 #include <netinet6/nd6.h> 73 #endif 74 #include <netinet/tcp.h> 75 #include <netinet/tcp_fsm.h> 76 #include <netinet/tcp_seq.h> 77 #include <netinet/tcp_timer.h> 78 #include <netinet/tcp_var.h> 79 #include <netinet/tcpip.h> 80 81 #include <netinet/udp.h> 82 #include <netinet/udp_var.h> 83 84 #include <netipsec/ipsec_support.h> 85 86 #include <machine/in_cksum.h> 87 88 #include <security/mac/mac_framework.h> 89 90 VNET_DEFINE_STATIC(bool, nolocaltimewait) = true; 91 #define V_nolocaltimewait VNET(nolocaltimewait) 92 SYSCTL_BOOL(_net_inet_tcp, OID_AUTO, nolocaltimewait, 93 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nolocaltimewait), true, 94 "Do not create TCP TIME_WAIT state for local connections"); 95 96 /* 97 * Move a TCP connection into TIME_WAIT state. 98 * inp is locked, and is unlocked before returning. 99 * 100 * This function used to free tcpcb and allocate a compressed TCP time-wait 101 * structure tcptw. This served well for 20 years but is no longer relevant 102 * on modern machines in the modern internet. However, the function remains 103 * so that TCP stacks require less modification and we don't burn the bridge 104 * to go back to using compressed time-wait. 105 */ 106 void 107 tcp_twstart(struct tcpcb *tp) 108 { 109 struct inpcb *inp = tptoinpcb(tp); 110 #ifdef INET6 111 bool isipv6 = inp->inp_inc.inc_flags & INC_ISIPV6; 112 #endif 113 114 NET_EPOCH_ASSERT(); 115 INP_WLOCK_ASSERT(inp); 116 117 /* A dropped inp should never transition to TIME_WAIT state. */ 118 KASSERT((inp->inp_flags & INP_DROPPED) == 0, ("tcp_twstart: " 119 "(inp->inp_flags & INP_DROPPED) != 0")); 120 121 tcp_state_change(tp, TCPS_TIME_WAIT); 122 soisdisconnected(inp->inp_socket); 123 124 if (tp->t_flags & TF_ACKNOW) 125 tcp_output(tp); 126 127 if (V_nolocaltimewait && ( 128 #ifdef INET6 129 isipv6 ? in6_localaddr(&inp->in6p_faddr) : 130 #endif 131 #ifdef INET 132 in_localip(inp->inp_faddr) 133 #else 134 false 135 #endif 136 )) { 137 if ((tp = tcp_close(tp)) != NULL) 138 INP_WUNLOCK(inp); 139 return; 140 } 141 142 tcp_timer_activate(tp, TT_2MSL, 2 * V_tcp_msl); 143 INP_WUNLOCK(inp); 144 } 145 146 /* 147 * Returns true if the TIME_WAIT state was killed and we should start over, 148 * looking for a pcb in the listen state. Otherwise returns false and frees 149 * the mbuf. 150 * 151 * For pure SYN-segments the PCB shall be read-locked and the tcpopt pointer 152 * may be NULL. For the rest write-lock and valid tcpopt. 153 */ 154 bool 155 tcp_twcheck(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th, 156 struct mbuf *m, int tlen) 157 { 158 struct tcpcb *tp = intotcpcb(inp); 159 char *s; 160 int thflags; 161 tcp_seq seq; 162 163 NET_EPOCH_ASSERT(); 164 INP_LOCK_ASSERT(inp); 165 166 thflags = tcp_get_flags(th); 167 #ifdef INVARIANTS 168 if ((thflags & (TH_SYN | TH_ACK)) == TH_SYN) 169 INP_RLOCK_ASSERT(inp); 170 else { 171 INP_WLOCK_ASSERT(inp); 172 KASSERT(to != NULL, 173 ("%s: called without options on a non-SYN segment", 174 __func__)); 175 } 176 #endif 177 178 /* 179 * NOTE: for FIN_WAIT_2 (to be added later), 180 * must validate sequence number before accepting RST 181 */ 182 183 /* 184 * If the segment contains RST: 185 * Drop the segment - see Stevens, vol. 2, p. 964 and 186 * RFC 1337. 187 */ 188 if (thflags & TH_RST) 189 goto drop; 190 191 #if 0 192 /* PAWS not needed at the moment */ 193 /* 194 * RFC 1323 PAWS: If we have a timestamp reply on this segment 195 * and it's less than ts_recent, drop it. 196 */ 197 if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && 198 TSTMP_LT(to.to_tsval, tp->ts_recent)) { 199 if ((thflags & TH_ACK) == 0) 200 goto drop; 201 goto ack; 202 } 203 /* 204 * ts_recent is never updated because we never accept new segments. 205 */ 206 #endif 207 208 /* Honor the drop_synfin sysctl variable. */ 209 if ((thflags & TH_SYN) && (thflags & TH_FIN) && V_drop_synfin) { 210 if ((s = tcp_log_addrs(&inp->inp_inc, th, NULL, NULL))) { 211 log(LOG_DEBUG, "%s; %s: " 212 "SYN|FIN segment ignored (based on " 213 "sysctl setting)\n", s, __func__); 214 free(s, M_TCPLOG); 215 } 216 goto drop; 217 } 218 219 /* 220 * If a new connection request is received 221 * while in TIME_WAIT, drop the old connection 222 * and start over if the sequence numbers 223 * are above the previous ones. 224 * Allow UDP port number changes in this case. 225 */ 226 if (((thflags & (TH_SYN | TH_ACK)) == TH_SYN) && 227 SEQ_GT(th->th_seq, tp->rcv_nxt)) { 228 /* 229 * In case we can't upgrade our lock just pretend we have 230 * lost this packet. 231 */ 232 if (INP_TRY_UPGRADE(inp) == 0) 233 goto drop; 234 if ((tp = tcp_close(tp)) != NULL) 235 INP_WUNLOCK(inp); 236 TCPSTAT_INC(tcps_tw_recycles); 237 return (true); 238 } 239 240 /* 241 * Send RST if UDP port numbers don't match 242 */ 243 if (tp->t_port != m->m_pkthdr.tcp_tun_port) { 244 if (tcp_get_flags(th) & TH_ACK) { 245 tcp_respond(tp, mtod(m, void *), th, m, 246 (tcp_seq)0, th->th_ack, TH_RST); 247 } else { 248 if (tcp_get_flags(th) & TH_SYN) 249 tlen++; 250 if (tcp_get_flags(th) & TH_FIN) 251 tlen++; 252 tcp_respond(tp, mtod(m, void *), th, m, 253 th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK); 254 } 255 INP_UNLOCK(inp); 256 TCPSTAT_INC(tcps_tw_resets); 257 return (false); 258 } 259 260 /* 261 * Drop the segment if it does not contain an ACK. 262 */ 263 if ((thflags & TH_ACK) == 0) 264 goto drop; 265 266 INP_WLOCK_ASSERT(inp); 267 268 /* 269 * If timestamps were negotiated during SYN/ACK and a 270 * segment without a timestamp is received, silently drop 271 * the segment, unless the missing timestamps are tolerated. 272 * See section 3.2 of RFC 7323. 273 */ 274 if (((to->to_flags & TOF_TS) == 0) && (tp->ts_recent != 0) && 275 (V_tcp_tolerate_missing_ts == 0)) { 276 goto drop; 277 } 278 279 /* 280 * Reset the 2MSL timer if this is a duplicate FIN. 281 */ 282 if (thflags & TH_FIN) { 283 seq = th->th_seq + tlen + (thflags & TH_SYN ? 1 : 0); 284 if (seq + 1 == tp->rcv_nxt) 285 tcp_timer_activate(tp, TT_2MSL, 2 * V_tcp_msl); 286 } 287 288 /* 289 * Acknowledge the segment if it has data or is not a duplicate ACK. 290 */ 291 if (thflags != TH_ACK || tlen != 0 || 292 th->th_seq != tp->rcv_nxt || th->th_ack != tp->snd_nxt) { 293 TCP_PROBE5(receive, NULL, NULL, m, NULL, th); 294 tcp_respond(tp, mtod(m, void *), th, m, tp->rcv_nxt, 295 tp->snd_nxt, TH_ACK); 296 INP_UNLOCK(inp); 297 TCPSTAT_INC(tcps_tw_responds); 298 return (false); 299 } 300 drop: 301 TCP_PROBE5(receive, NULL, NULL, m, NULL, th); 302 INP_UNLOCK(inp); 303 m_freem(m); 304 return (false); 305 } 306